Run test_07_22 only on http/1.1 and h2. For h3, we get unreliable tests
results in CI due to (probably) nghttpx not always reporting the RESET
from the backend and the test timing out.
Closes#21286
- It is slower and uses more memory than the alternatives and is only
experimental in curl.
- We disable a few tests for OpenSSL-QUIC because of flakiness
- It gets little attention from OpenSSL and we have no expectation of the
major flaws getting corrected anytime soon.
- No one has spoken up for keeping it
- curl users building with vanilla OpenSSL can still use QUIC through the
means of ngtcp2
Closes#20226
Do not generate a 400 response code, but use a 200 one. The upload needs
to fail on sending, not on seeing a 400 response. Seeing a 400 before
the sending fails (when CI timings shift) will expose the wrong error
code.
Ref: #20112Closes#20164
We recently allowed a larger send buffer in ngtcp2 streams. This allowed
curl to send more early data then previously when the server was slow in
performing the handshake. This led to flaky test failures when the
amount of early data was larger than expected.
Change test expectations to allow for varying amount of early data.
Ref: #20112Closes#20161
Description of how this works in `docs/internal/RATELIMITS.ms`.
Notable implementation changes:
- KEEP_SEND_PAUSE/KEEP_SEND_HOLD and KEEP_RECV_PAUSE/KEEP_RECV_HOLD
no longer exist. Pausing is down via blocked the new rlimits.
- KEEP_SEND_TIMED no longer exists. Pausing "100-continue" transfers
is done in the new `Curl_http_perform_pollset()` method.
- HTTP/2 rate limiting implemented via window updates. When
transfer initiaiting connection has a ratelimit, adjust the
initial window size
- HTTP/3 ngtcp2 rate limitin implemnented via ack updates
- HTTP/3 quiche does not seem to support this via its API
- the default progress-meter has been improved for accuracy
in "current speed" results.
pytest speed tests have been improved.
Closes#19384
- fix test_17_20 flakiness: the test case did not have `nghttpx` in
its parameters, causing it to no check if a reload was necessary.
When that test ran behind one that gave nghttpx another certificate,
eg. in parallel mode, it used the wrong pinned pubkey.
- Have `env` provide lists of HTTP protocol versions available for
testing. Replace parameterized tests on a fixed protocol list with
the dynamic one from env. This makes checks for protocol availability
in the test function bodies superfluous.
refs #19489Closes#19540
The HTTP/3 tests did send 20 transfers against nghttpx with a backend
that failed the uploads with a 400 and an incomplete response body. This
causes stream resets.
Apache keeps the connection open, but newer nghttpx closes the front
connection after "too many" reset. When that bites, it depends on the
number of transfers ongoing how the test case fails. This led to flaky
outcomes.
Reduce the transfers to just a single one and check the result of
that one. Parallelism is not important here.
refs #19489Closes#19530
To allow running pytests on more curl configurations.
Also delete a redundant H3 feature check from test_17_14_expired_cert.
Cherry-picked from #19407Closes#19412
Change the test certificate to carry a altname 'dns:127.0.0.1' which
should *not* match in test_17_05_bad_ip_addr.
wolfSSL: since `wolfSSL_check_domain_name()` does not differentiate
between DNS and IP names, use if only for DNS names. For IP addresses,
get the peer certificate after the handshake and check that using
wolfSSL_X509_check_ip_asc().
Unfortunately, this succeeds where it should not, as wolfSSL internally
used the same check code for both cases. So, skip the test case until
wolfSSL fixes that.
Reported-by: Joshua Rogers
Closes#19252
- configure/cmake support for enabling the option
- supported in OpenSSL and GnuTLS backends
- when configured, Apple SecTrust is the default trust store
for peer verification. When one of the CURLOPT_* for adding
certificates is used, that default does not apply.
- add documentation of build options and SSL use
Closes#18703
libtests and clients were built the same way after recent overhauls.
libtests are used by runtests, clients by pytests.
Merge clients into libtests, aligning their entry function signature,
dropping common utility functions, and simplifying the build.
Note: After this patch `CURLDEBUG` applies to cli tests, when enabled.
Also:
- lib552: drop local copy-paste debug callback in favor of testtrace.
- lib552: drop local copy-paste dump function in favor of testtrace.
- clients: use `long` for HTTP version, drop casts.
- clients: replace local dump function in favor of testrace clone.
- sync cli test entry function prototype with libtests'.
- h2_serverpush: replace local trace callback with testtrace.
- de-duplicate 3 websocket close, ping, ping, functions. Kept the pong
iteration from `ws_pingpong`. Note: the pong clone in `lib2304` was
returning an error when `curl_ws_recv()` returned non-zero and
the payload matched the expected one anyway. After this patch, this
case returns success, as it does in `ws_pingpong`.
`lib2304` keeps passing, but I'm not sure if the previous behavior
was intentional.
- display full value in websocket close, ping, pong, drop casts.
Closes#18079
To make building the http client tests faster, with no duplication, by
using the build method that other test binaries already use.
The difference compared to other tests is that these don't use internal
libcurl headers or code. With the exception of `curl_config.h`, for
a feature macro.
Before this patch, these tests were built like examples.
Also:
- de-duplicate code and give unique names to colliding symbols.
- add local getopt implementation and enable all code for MSVC.
Adapted for curl via Public Domain source:
4e618ef782/getopt.h
Credits-to: Christopher Wellons
Thanks!
Closes#17627
Since the amount of early data sent in the upload test_07_70 varies
much with how fast the server respondes and completes the handshake,
assert that we did sent at least *some* early data instead of relying
on a specific amount.
Closes#17575
When inspecting a possible follow HTTP request, the result of a rewind
of the upload data was ignored as it was not clear at that point in time
if the request would become a GET.
This initiated the followup, rewound again, which failed again and
terminated the follow up.
This was confusing to users as it was not clear of the follow up was
done or not.
Fix: fail the early rewind when the request is not converted to GET.
Fixes#17472Closes#17474
Reported-by: Jeroen Ooms
They take about 4.5 minutes of CI time in GHA/macos.
Also:
- autotools: improve `caddy`, `vsftpd` detection.
Bringing it closer to cmake.
- autotools: fix `--with-test-caddy=no`, `--with-test-vsftps=no`,
`--with-test-nghttpx=no` options.
- cmake: sync `nghttpx` default with autotools.
- pytest: disable failing mbedTLS tests on macOS CI.
- pytest: disable failing earlydata tests on macOS CI.
- GHA/macos: keep vsftpd pytests disabled due to lengthy run times.
- pytest: fix test_05_04 for LibreSSL. Ref: #17367
Authored-by: Stefan Eissing
Remaining issues:
- some unidentified tests taking a long time with mbedTLS:
`================= 462 passed, 278 skipped in 347.93s (0:05:47) =================`
Ref: https://github.com/curl/curl/actions/runs/15073354301/job/42374999041#step:17:1536
Workaround: not enabling pytest for mbedTLS jobs
- 17 FTP tests taking a long time (affecting all TLS backends):
without vsftpd:
`====================== 496 passed, 244 skipped in 56.15s =======================`
Ref: https://github.com/curl/curl/actions/runs/15073354301/job/42374998230#step:17:1536
with vsftpd:
`================= 513 passed, 227 skipped in 409.37s (0:06:49) =================`
Ref: https://github.com/curl/curl/actions/runs/15073678568/job/42376039672?pr=17362#step:17:1537
Workaround: force-disable vsftpd.
- 100 tests failing with SecureTransport. Let's ignore that due to imminent deprecation.
Ref: https://github.com/curl/curl/actions/runs/15055652333/job/42320873732#step:17:15362
Follow-up to 30ef79ed93#17295
Follow-up to 9147903366#16518Closes#17362
Require now pytest-xdist from tests/http/requirements.txt and
run pytest in 'auto' parallel mode (counts cpu cores).
For CI runs, set the worker count to 4, overriding the
core count of 2 exposed in the images.
- use Filelock to generate allocated ports at start for all
workers and have subsequent workers just read the file and
take the ports for their slot
- make httpd config clearing a function fixture so every test
starts with a clean httpd config
- have fixture `configures_httpd` as parameter of test cases
that configure httpd anyway, saving one reload
- add pytest-xdist and filelock to required pyhton modules
- add installs to ruff CI
- give live checks waiting for a server to start up longer time
- add fixtures to tests that rely on a server
- do not stop servers unnecessarily. failures may not start them
properly again, leading to unexpected fails in whatever follows
- add a https: port to httpd that is *not* back by QUIC to allow
failover tests without stopping the QUIC server
Closes#17295
Depending on timing when the server aborting the connection is detected,
the reported curl exit code may vary. Check for the possible set of
expected codes instead of a single one.
Closes#17083
Change multi's book keeping of transfers to no longer use lists, but a
special table and bitsets for unsigned int values.
`multi-xfers` is the `uint_tbl` where `multi_add_handle()` inserts a new
transfer which assigns it a unique identifier `mid`. Use bitsets to keep
track of transfers that are in state "process" or "pending" or
"msgsent".
Use sparse bitsets to replace `conn->easyq` and event handlings tracking
of transfers per socket. Instead of pointers, keep the mids involved.
Provide base data structures and document them in docs/internal:
* `uint_tbl`: a table of transfers with `mid` as lookup key,
handing out a mid for adds between 0 - capacity.
* `uint_bset`: a bitset keeping unsigned ints from 0 - capacity.
* `uint_spbset`: a sparse bitset for keeping a small number of
unsigned int values
* `uint_hash`: for associating `mid`s with a pointer.
This makes the `mid` the recommended way to refer to transfers inside
the same multi without risk of running into a UAF.
Modifying table and bitsets is safe while iterating over them. Overall
memory requirements are lower as with the double linked list apprach.
Closes#16761
When the server sends HEADER/CONTINUATION frames that exceed nghttp2's
size, this error is being reported via the on_invalid_frame_recv
callback. Without registering there, it will go unnoticed.
RST the stream when such a frame is encountered.
Closes#16544
Enable TLS Early Data for wolfSSL:
- merge WOLFSSL_CTX and WOLFSSL setup from ngtcp2 with the general
implemenation in wolfssl.c
- enable for QUIC via ngtcp2
- give Curl_vquic_tls_init() a `struct alpn_spec` like used for the TCP
case. Adapt gnutls and other users.
- enable pytest test cases for early data with wolfSSL
and while this messes up wolfssl.c anyway, do
- rename all struct/functions with prefix 'wolfssl_' to 'wssl_' to not
pollute that name prefix
- rename `ctx/handle` to `ssl_ctx/ssl`, as used in openssl case
Closes#16167
When a QUIC TLS session announced early data support and
'CURLSSLOPT_EARLYDATA' is set for the transfer, send initial request and
body (up to the 128k we buffer) as 0RTT when curl is built with
ngtcp2+gnutls.
QUIC 0RTT needs not only the TLS session but the QUIC transport
paramters as well. Store those and the earlydata max value together with
the session in the cache.
Add test case for h3 use of this. Enable quic early data in nghttpx for
testing.
Closes#15667
The base64 mime encoder stalls when it cannot encode a full 3 byte input
set into the read buffer. The workaround for this limitation was
incomplete and could lead to stalled transfers when the last chunk to
upload was smaller than 4 bytes.
Use a tmp buffer on small reads to allow mime encoders more space to put
their things.
Add test case reproducing the issue and fix.
Reported-by: Alexis Savin
Fixes#15688Closes#15691
When a server signals EOS from its side and the curl upload is
unfinished and the server has not given a positive HTTP status response,
auto RST the stream to signal that the upload is incomplete and that the
whole transfer can be stopped.
Fixes the case where the server responds with 413 on an upload but does
not RST the stream from its side, as httpd and others do.
Reported-by: jkamp-aws on github
Fixes#15316Closes#15325
Based on #14135, implement TLSv1.3 earlydata support for the curl
command line, libcurl and its implementation in GnuTLS.
If a known TLS session announces early data support, and the feature is
enabled *and* it is not a "connect-only" transfer, delay the TLS
handshake until the first request is being sent.
- Add --tls-earldata as new boolean command line option for curl.
- Add CURLSSLOPT_EARLYDATA to libcurl to enable use of the feature.
- Add CURLINFO_EARLYDATA_SENT_T to libcurl, reporting the amount of
bytes sent and accepted/rejected by the server.
Implementation details:
- store the ALPN protocol selected at the SSL session.
- When reusing the session and enabling earlydata, use exactly
that ALPN protocol for negoptiation with the server. When the
sessions ALPN does not match the connections ALPN, earlydata
will not be enabled.
- Check that the server selected the correct ALPN protocol for
an earlydata connect. If the server does not confirm or reports
something different, the connect fails.
- HTTP/2: delay sending the initial SETTINGS frames during connect,
if not connect-only.
Verification:
- add test_02_32 to verify earlydata GET with nghttpx.
- add test_07_70 to verify earlydata PUT with nghttpx.
- add support in 'hx-download', 'hx-upload' clients for the feature
Assisted-by: ad-chaos on github
Closes#15211
Due to timings on paused response receive, the error code may vary due
to the location where it is detected that the server closed the transfer
prematurely.
Be more lenient in always allowing PARTIAL_FILE as ok.
Closes#15138
When uploading data from stdin ('-T -'), and the EOS was only detected
on a 0-length read, the EOS was not forwarded to the filters. This led
HTTP/2 to hang on not forwarding this to the server.
Added test_07_14 to reproduce and verify.
Fixes#14870
Reported-by: nekopsykose on github
Closes#14877
- replace the counting of upload lengths with the new eos send flag
- improve frequency of stream draining to happen less on events where it
is not needed
- this PR is based on #14220
http2, cf-h2-proxy: fix EAGAINed out buffer
- in adjust pollset and shutdown handling, a non-empty `ctx->outbufq`
must trigger send polling, irregardless of http/2 flow control
- in http2, fix retry handling of blocked GOAWAY frame
test case improvement:
- let client 'upload-pausing' handle http versions
Closes#14253
- add 2 variations on test_07_42 which PAUSEs uploads
and response connections terminating either right away
or after the 100-continue response
- when detecting the connection being closed in transfer.c
readwrite_data(), clear ALL send bits in data->req.keepon.
It no longer makes send to wait for a KEEP_SEND_PAUSE or HOLD.
- in the protocol client writer add the check for incomplete
response bodies. When an EOS is seen and the length is known,
check that and fail if bytes are missing.
Reported-by: Sergey Bronnikov
Fixes#13740Closes#13750
A transfer with a completed download that is still uploading needs to
check the connection state when it is PAUSEd, since connection
close/errors would otherwise go unnoticed.
Reported-by: Sergey Bronnikov
Fixes#13260Closes#13271
Move all handling of HTTP's `Expect: 100-continue` feature into a client
reader. Add sending flag `KEEP_SEND_TIMED` that triggers transfer
sending on general events like a timer.
HTTP installs a `CURL_CR_PROTOCOL` reader when announcing `Expect:
100-continue`. That reader works as follows:
- on first invocation, records time, starts the `EXPIRE_100_TIMEOUT`
timer, disables `KEEP_SEND`, enables `KEEP_SEND_TIMER` and returns 0,
eos=FALSE like a paused upload.
- on subsequent invocation it checks if the timer has expired. If so, it
enables `KEEP_SEND` and switches to passing through reads to the
underlying readers.
Transfer handling's `readwrite()` will be invoked when a timer expires
(like `EXPIRE_100_TIMEOUT`) or when data from the server arrives. Seeing
`KEEP_SEND_TIMER`, it will try to upload more data, which triggers
reading from the client readers again. Which then may lead to a new
pausing or cause the upload to start.
Flags and timestamps connected to this have been moved from
`SingleRequest` into the reader's context.
Closes#13110
- add test cases for rate limiting uploads for all
http versions
- fix transfer loop handling of limits. Signal a re-receive
attempt only on exhausting maxloops without an EAGAIN
- fix `data->state.selectbits` forcing re-receive to also
set re-sending when transfer is doing this.
Reported-by: Karthikdasari0423 on github
Fixes#12559Closes#12586
- there seems to be a code path that cleans up easy handles without
triggering DONE or DETACH events to the connection filters. This
would explain wh nghttp2 still holds stream user data
- add GOOD check to easy handle used in on_close_callback to
prevent crashes, ASSERTs in debug builds.
- NULL the stream user data early before submitting RST
- add checks in on_stream_close() to identify UNGOOD easy handles
Reported-by: Hans-Christian Egtvedt
Fixes#10936Closes#12562
- refs #11342 where errors with git https interactions
were observed
- problem was caused by 1st sends of size larger than 64KB
which resulted in later retries of 64KB only
- limit sending of 1st block to 64KB
- adjust h2/h3 filters to cope with parsing the HTTP/1.1
formatted request in chunks
- introducing Curl_nwrite() as companion to Curl_write()
for the many cases where the sockindex is already known
Fixes#11342 (again)
Closes#11803
